First, do no harm:
The Team
Towards Risk Analysis and Risk Management
in Radiation Therapy
Robert C. Lee
Edidiong Ekaette
Peter Dunscombe, Ph.D.
University of Calgary/Tom Baker Cancer Centre
Karie-Lynn Kelly
David Cooke
Professional Symposium: Medical Errors and Medical Physics
Chris Newcomb
AAPM Annual Meeting, Seattle, July 2005
Peter Craighead
Outline
Acknowledgement
1. The Issues
Canadian Institutes of Health
Health Quality Council of Alberta
2. Towards Quantifying Risk in Radiation
Therapy: Risk Analysis
3. Towards Integrating Operations with Risk
Analysis: Risk Management
4. Conclusions
1
Issues
Issues
The Issues
The Issues
Patients are entitled to hold the expectation
that they will not be harmed during an
encounter with a health care system.
Health care providers have an obligation
to minimize the risk of harm to patients
(and to be seen to be doing so).
Issues
Issues
The Issues
The US Institute of Medicine estimates
that between 44,000 and 98,000 deaths
per year in the US are attributable to
medical errors.
To Err is Human: Building a Safer Health System. 1999. IoM, National Academy Press, Washington, DC
The Issues
In Canadian hospitals, the overall
incidence rate of adverse events was
estimated to be 7.5%. 37% of these were
preventable and 20% resulted in death.
Journal of the Canadian Medical Association. 2004, 170, 1678-1686
2
Issues
Issues
Premise
Benefit and Harm
We can meet patients’ expectations and providers’
obligations by
maximizing benefit to patients
•
minimizing harm to patients
Harm
Harm
Benefit
Frequency
•
while, at the same time,
•
staying within resource constraints
•
acknowledging other stakeholders
•
we quantitatively understand the characteristics of the
systems within which patients and providers interact.
-
+
Deviation from optimum dose
if
Harm results from an unacceptable deviation from a
“quality treatment”
Issues
Issues
Benefit and Harm
Risk Descriptors 1
An acceptable treatment can be
accepted without further
improvement.
Harm
An unacceptable treatment is one
that would not be acceptable on
any reasonable basis.
(adapted from ICRP 60)
Harm
Incident: an unwanted or unexpected change from normal
system behaviour that causes, or has the potential to
cause, an adverse event
+
Adverse Event: an incident that occurs during the process
of providing health care and results in sub-optimal
treatment, unintended injury or complication leading to
disability, death or prolonged hospital stay for the
patient
Benefit
Frequency
A tolerable treatment is not ideal
but can reasonably be tolerated.
Deviation from optimum dose
• As Close As Reasonably Achievable to the Optimum: quality management
• Avoidance of Harm: risk management
3
Issues
Issues
Risk Descriptors 2
Risk Defined
Error: failure to complete a planned action as it was
intended or a situation in which incorrect methods
and/or data are used in an attempt to achieve a given aim
(relates to rules and skills)
r = f(p,c,)
where:
r is the risk
p is the probability of the event
Mistake: selection of inappropriate plans and decisions
(relates to knowledge)
c is the consequence of the event per patient
Misadministration: deviation from a prescription that
exceeds a defined value
The number of patients involved might also be a factor in
assessing risk.
Issues
Risk Analysis
Current Approaches:
Comparison of Alternatives
Risk Analysis
Method
Criterion
Ad hoc
Descriptive
statistics
RCA
FMEA
Microsystems
modeling
Systematic
-
+
+
+
-
-
-
+
-
+
+
System-oriented
-
-
+
+
+
+
+
Quantitative
-
+
-
Semi +
Semi +
+
+
Addresses parameter
uncertainty
-
Semi +
-
-
-
+
+
Addresses variability
-
+
-
-
-
+
+
Can model rare incidents
-
-
-
-
-
+
+
Addresses multiple
objectives
-
-
-
-
-
-
+
Addresses resource
constraints
-
-
-
-
-
-
+
Addresses changes to the
system
-
-
-
-
+
Decision
analysis
Proactive/predictive
-
+
PRA
-
To quantify risk to a patient effectively we need to know two
things about modes of system failure.
+
+
We need to know:
•
the probability of a particular event occurring, p
•
and the consequence if that event does occur, c
•
The number of patients actually or potentially affected is
likely to influence risk management strategies.
r = f(p,c)
HTA Initiative #15. Alberta Heritage Foundation for Medical Research. www.ahfmr.ab.ca
4
Risk Analysis
Risk Analysis
Process Map
Fault Tree
•
If we can develop a fault tree we should be able to work out p
•
Development of a fault tree will be facilitated by a
taxonomy
•
Development of a taxonomy will be facilitated by a process
map
r = f(p,c)
Thomadsen et al. 2003 IJROBP 57, 1492-1508
Chang et al. 2005 IJQHC 17, 95-105
Risk Analysis
Risk Analysis
Taxonomy: Domains
The Process Map suggests that there are five domains within
which Incidents may occur:
1. Assessment
2. Prescription
3. Treatment Preparation
4. Treatment Delivery
5. Follow-up
Taxonomy: Prescription Elements
Should an Incident occur it will have consequences for
one or more Prescription Elements
In radiation therapy, these Prescription Elements can
conveniently be classified as:
• Dose
• Volume
both are 4D concepts
and can be applied to targets and organs at risk
5
Risk Analysis
Risk Analysis
Taxonomy: Source
Taxonomy: Occurrence
We consider that Incidents may arise from two general
sources:
We consider that Incidents may occur in two general
ways:
• Infrastructure – calibrated equipment such as linacs;
software; hard and soft copy beam data; protocols; etc.
• Sporadic – will occur unpredictably
• Processes – performing the required tasks in the
environment created by the infrastructure provided
A Digression
• Systematic – will occur predictably under similar
circumstances
A Digression
A Digression: Uncertainties in
Radiation Therapy
What about random errors?
We consider what are frequently called random errors, for
example in connection with patient set-up, to be
uncertainties in the sense that they are associated with the
limitations of measurement techniques and the ability of
humans to repeat a process in exactly the same way.
A Digression: Uncertainties in
Radiation Therapy
We have undertaken a study of the probability of a woman
with breast cancer being assigned to the inappropriate
treatment technique due uncertainties in staging.
A Bayesian approach was followed and included staging
based on imaging, pathology and physical examination.
Lee et al. Medical Decision Making, to be published
6
A Digression
Risk Analysis
Simulation results for RT
treatment technique
prescription
Taxonomy
Assessment
Required
true
treatment
‘none’
‘2-field’
‘4-field’
‘palliative’
none
100.0
0.0
0.0
0.0
2-field
0.0
98.3
0.0
0.0
4-field
0.0
1.5
92.3
1.9
palliative
0.0
0.2
7.7
98.1
Prescription
Preparation
Delivery
Follow up
Prescribed treatment
Volume
Dose
Process
Infrastructure
Sporadic
Process
Infrastructure
Systematic
NOTE: Results are mean percentages.
Risk Analysis
Risk Analysis
Analysis of existing databases
As a first step towards a fault tree for radiation therapy we
have taken the incidents reported in:
1. NRC database
2. IAEA “Lessons learned from accidental exposures in
radiotherapy” 2000
3. ROSIS database
And categorized them according to the taxonomy
7
Risk Analysis
Risk Analysis
Reported Incidents by
Prescription Element
Reported Incidents by Domain
0.9
0.6
% of reported incidents
% of reported incidents
0.7
0.5
ROSIS
0.4
NRC
0.3
IAEA
0.2
0.1
0
Prescription
Preparation
Treatment
0.8
0.7
0.6
ROSIS
0.5
NRC
0.4
IAEA
0.3
0.2
0.1
0
Dom ain
Dose
Volume
Prescription elem ents
Risk Analysis
Risk Analysis
Reported Incidents by Source
Reported Incidents by Occurrence
1.2
1
0.8
ROSIS
0.6
NRC
IAEA
0.4
0.2
% of reported incidents
% of reported incidents
1.2
1
0.8
ROSIS
0.6
NRC
IAEA
0.4
0.2
0
0
Process
Infrastructure
Source
Systematic
Sporadic
Nature of occurre nce
8
Risk Analysis
Risk Analysis
Consequence Metrics
So what do we know?
There is some, but not a high, degree of consistency
between the relative incident rates between the different
databases.
As the databases were established for different purposes
this is not too surprising.
However, the lessons to be learnt from conflicting
information are unclear.
Risk Analysis
With every error will associated a consequence denoted
by c
r = f(p,c)
Metrics that have been used include:
• Death
• Disability
• Unexpected hospitalization
• Life years lost
• Quality adjusted life years lost
Risk Analysis
Consequence Metrics
In radiation therapy, prescription elements have two
components: dose and volume; both are 4D concepts.
EUD as a consequence metric, c
Standard two field tangential breast treatment
EUD Breast
EUD Lung
Prescribed Treatment
51.0
12.8
15MV instead of 6MV – whole course
55.7
14.0
SAD + 10cm – whole course
42.3
10.7
SAD + 10cm – one fraction
50.6
12.7
Missing wedge - whole course
55.1
12.9
Missing wedge – one fraction
51.1
12.8
A biological EUD can also be calculated – 4D consequence
metric
Coll. Rotation 100 – whole course
50.6
12.7
Coll. Rotation 100 – one fraction
51.0
12.7
Niemerko 1997 Med Phys 24, 103-110
With thanks to Sandy Iftody C.M.D. and Nicolas Ploquin M.Sc.
Is there a way of combining these two elements into one
metric which can be used to quantify the consequence?
The Equivalent Uniform Dose consolidates 3 dimensional
dose and volume information into one number.
EUD
1
N
1/ a
a
i
D
i
Treatment Configuration
9
Risk Analysis
Risk Management
EUD as a consequence metric, c
The Equivalent Uniform Dose may be an
appropriate measure of the consequence of an
incident in radiation therapy.
The EUD allows us to compare both dose and
volume effects using a metric that should be
indicative of outcome.
Integrating Operations with Risk
Analysis: Risk Management
There are three components of operations that could
readily be coordinated with risk analysis:
1. Assignment of responsibilities
2. Quality control
3. Incident learning
Risk Management
Risk Management
Integrating Operations with Risk
Analysis: Responsibilities
Domain
Assessment
Responsibilities in conventional external beam radiation therapy
Procedures
Primary Responsibility
Infrastructure
Process
Diagnosis
Radiation oncologist
Staging
Prescription
Preparation
Delivery
Follow-up
Pathologist
Dose and Volume
Treatment aids
Treatment planning
Physicist
Physicist
Physicist
Radiation oncologist
Mould room staff
Dosimetrist
Simulation
Physicist
Simulator therapist
Treatment delivery
Verification
Physicist
Physicist
Integrating Operations with Risk
Analysis: Responsibilities
On the basis of a risk analysis and by being very
specific about responsibilities of the different
members of the radiation therapy team we can
direct limited continuing education resources to
those areas where risk is the highest.
Delivery therapist
Radiation oncologist
Delivery therapist
Radiation oncologist
10
Risk Management
Risk Management
Integrating Operations with Risk Analysis:
Quality Control
Monthly
ML1
ML2
ML3
ML4
ML5
ML6
ML7
ML8
ML9
ML10
ML11
ML12
ML13
ML14
ML15
ML16
ML17
Emergency off
Wedge, tray cone interlocks
Accessories integrity and centering
Gantry angle readouts
Collimator angle readouts
Couch position readouts
Couch isocentre
Couch angle
Optical distance indicator
Crosswire centering
Light/radiation coincidence
Field size indicator
Relative dosimetry
Central axis depth dose reproducibility
Beam flatness
Beam symmetry
Records
Functional
Functional
Functional
0.5
1
0.5
1
1
2
1
2
0.5o
1o
1
2
1
2
1
2
1
2
1%
2%
1%/2mm
2%/3mm
2%
3%
2%
3%
Complete
Integrating Operations with Risk
Analysis: Quality Assurance
By linking quality assurance to a formal risk analysis we
can direct limited resources to those areas where the risk
is highest. (Evidence Based Quality Assurance)
For example, certain high risk activities may be
subjected to more intense verification than others.
Risk Management
Risk Management
Integrating Operations with Risk
Analysis: Incident Learning
Integrating Operations with Risk
Analysis: Incident Learning
Incident Characteristics (CLINICAL)
Impact:
# patients affected: _______________________
# fractions per patient affected:______________
# fields per fraction affected: ________________
Deviation from prescribed dose: ______________
Deviation from prescribed volume: ____________
Dosimetrist/medical physicist who analyzed incident:
Name:__________________ Date: ___________
Signature: _______________________________
Name:__________________ Date: ___________
Signature: _______________________________
Domain:
For clinical incidents:
Domain
Assessment
Discovery of
incident
Origin of
incident
By linking incident learning to a formal risk
analysis we can develop and refine the risk
analysis to reflect local circumstances.
Prescription
Preparation
Treatment
Follow up
Other:
___________
11
Conclusions
Conclusions
Conclusions
1. Opportunities exist for improving the way we
collect and analyze reports of incidents in
radiation therapy.
2. With a better understanding of risk in radiation
therapy we will be better able to meet patients’
expectations and providers’ obligations.
Final Thought
Incidents are like photons
You can reduce the number
You can reduce the energy or impact
But you can never get rid of them all
12